High speed confocal 3d profilometer design, development, experimental results

Among the existing optical profilometry techniques, confocal technique is very special because its measurement resolution can be customized to be as small as 0.01µm while its measurement

HIGH SPEED CONFOCAL 3D PROFILOMETER:

DESIGN, DEVELOPMENT, EXPERIMENTAL RESULTS

ANG KAR TIEN

(M.Eng., NUS) (B.Eng (Hons.), UTM)

A THESIS SUBMITTED

FOR THE DEGREE OF DOCTOR OF PHILOSOPHY

DEPARTMENT OF ELECTRICAL

AND COMPUTER ENGINEERING

NATIONAL UNIVERSITY OF SINGAPORE

2014

Acknowledgements

First and foremost, my sincerest gratitude and thanks go to my research supervisors, Assoc Prof Arthur Tay from National University of Singapore (NUS), and Dr Fang Zhong Ping, from Singapore Institute of Manufacturing (SIMTech) Both of them have supported me throughout my research with their patience, knowledge, useful advice and resources Without their consistent involvements, encouragement, stimulating ideas, suggestions and help in every aspect of my research, this thesis would not have been completed

My research project is collaboration between NUS and SIMTech So, I would like to take this opportunity to express my gratitude to NUS for offering

me a PhD research scholarship and its excellent library services and other facilities and services Secondly, I would like to express my gratitude to SIMTech for giving me an opportunity to do this research and providing the financial support, tools and equipment

I would also like to say thank you to Dr He Wei (SIMTech research scientist), Sukresh Sivasailam and John Britto Montfort (NUS master students) for fabricating Nipkow disks Thanks to all of them who has ever helped in to complete my prototype I am also grateful to friendly and supportive SIMTech staffs, e.g Dr Zhang Ying, Dr Seck Hon Huen, Dr Li Xiang, Dr Yu Xia, Dr

Li Hao, Dr Chong Wee Keat, Dr Dr Xu Jian, Dr Isakov Dmitry, Mdm Xie Hong, Mdm Liu Yuchan, Ms Daphne Seah, Ms Liew Seaw Jia, Mr Ng

Khoon Leong, Mr Yong Hock Hung, Mdm Lee Yeng Lang, etc Thanks to all whom I have unintentionally left out, but give me a helping hands and friendly smile while I was doing research in SIMTech

Besides that, I also want to thank NUS laboratory technologists, Mdm

S Mainavathi, Mdm Ho Leng Joo, Mr Joseph Ng Gek Leng, Mr Tan Chee Siong, Mdm Aruchunan Sarasupathi, Mr Zhang Heng Wei, etc for their unconditional support I also like to thanks all my friends and colleagues who have shared inspiring experiences and entertainment moment with me: Dr Ngo Yit Sung, Dr Teh Siew Hong, Dr Qu Yifan, Dr Nie Maowen, Mr Yong See Wei, Mr Conan Toh, Mr Henry Tan, Dr Chua Ding Juan, Dr Yang Rui, etc Thanks to all whom I have unintentionally left out, but give me a helping hands and friendly smile while I was doing research in NUS

I also like to take this opportunity to thank my parents, sisters and brothers-in-law for their support and encouragement My cute nephew and nieces also bring me a lot of joys Last but not least, I would like to thank all

my friends, relatives, ex-class mates, and ex-colleagues e.g Dr Yeak Su Hoe,

Dr Claus Dusemund, Dr Ong Kean Leong, Mr C.P Ang, Mr Er Chin Hai, etc for their friendship, caring, and encouragement Many thanks to all whom

I have unintentionally left out

Contents

Acknowledgements i

Summary v

List of Tables viii

List of Figures ix

List of Abbreviations xiv

Chapter 1 1

Introduction 1

1.1.Motivations 1

1.2 Contributions 6

1.3 Organization 9

Chapter 2 10

Literature Review 10

2.1 Introduction 10

2.2 An Overview of Optical Profilometers 12

2.3 Point-wise Optical Techniques 13

2.3.1Triangulation 13

2.3.2 Confocal 14

2.3.3 Point Autofocus 15

2.4 Whole-Field Optical Techniques 17

2.4.1 Focus Variation 17

2.4.2 Phase Shifting Interferometry 19

2.4.3 Digital Holographic Microscopy 21

2.4.4 Coherence Scanning Interferometry 22

2.4.5 Pattern Projection Methods 24

2.5 Current Confocal Profilometry Technology 24

2.6 Summary 27

Chapter 3 28

Measurement of Topography using Confocal Microscope 28

3.1 Introduction 28

3.2 Experiment Setup 29

3.3 Calibration Process 32

3.4 Feature Height Extraction 34

3.5 Summary 41

Chapter 4 42

Design of Prototype 42

4.1 Introduction 42

4.2 The Illumination System 44

4.3 The Microscopy System 48

4.4 The Imaging System 49

4.5 The Spinning Disk System 53

4.6 Other Components 58

4.7 The Final System 59

4.8 Summary 65

Chapter 5 66

Enhanced System Development and Testing 66

5.1 Introduction 66

5.2 Vector Projection Technique 67

5.3 Various Height Retrieval Methods 76

5.4 The Range for Accurate Measurement 90

5.5 Height Retrieval Using Multiple Images 98

5.6 Summary 105

Chapter 6 107

Conclusion and Future Works 107

6.1 Conclusion 107

6.2 Future Works 109

Author’s Publications 114

Bibliography 116

Appendix A 127

Appendix B 143

Appendix C 148

Appendix D 151

Summary

Three-dimensional (3D) profilometer is a surface measurement instrument and it is a key metrology tool for many current state-of-art manufacturing industries Nowadays there is a high demand for high speed 3D profilometer in the field of precision engineering, micromachining, optoelectronic, electronic, photonic, optic, microfluidic, medical implant, material science, tribology, large area printing, etc

Among the existing optical profilometry techniques, confocal technique is very special because its measurement resolution can be customized to be as small as 0.01µm while its measurement range can be customized as large as 20mm Unlike interferometry techniques, confocal technique does not encounter phase wrapping problem Confocal technique does not suffer the drawbacks faced by triangulation and pattern projection such as occlusion, and multiple reflections Unlike focus variation technique, confocal technique can measure transparent surface In addition confocal technique can measure feature with discontinuity such as large step, and pillar Many existing commercial optical profilometers have difficulties to measure the 3D topography of miniature pillar structures of a transparent microfluidic device Confocal technique is very suitable to measure the pillar structures of the microfluidic device

However, the measurement times of the commercially available confocal profilometers are quite long Typically one measurement takes a few

minutes to several hours depends on total number of measurement points Since these confocal profilometers use confocal point sensor, mechanical scanning process slows down the measurement speed Thus, the usefulness of these confocal profilometers is thus limited due to its slow measurement speed

In this work, we have designed, and developed a high speed confocal 3D profilometer by combining the spinning Nipkow disk and chromatic confocal technique In this configuration, a color camera is used instead of spectrometer as the detector The confocal system needs to be calibrated for each sample material before it can be used for measurement During measurement, a confocal image of the sample is captured and the color information of each pixel is compared with the calibration data in order to determine the surface height of the pixel

Various height retrieval methods have been studied and compared The Vector Projection technique has been developed to replace the discrete point technique to improve the resolution of the measurement reading Multiple- image height retrieval scheme also has been developed to extend the measurement range Finally, the high speed confocal 3D profilometer prototype system is used to measure the surface topography of the pillar structures of a microfluidic device Experimental results demonstrate the feasibility and accuracy of the proposed approach The vertical resolution of the prototype system is about 0.05 µm The prototype system can measure the

surface topography of a sample with the size of 0.44mm × 0.33mm and the resolution of 1360 ×1024 pixels within 10 seconds

List of Tables

Table 5.1: The height of the sample, H1, retrieved using the RGB, HSV, XYZ,

XZ, and HS methods 83Table 5.2: The height of the sample, H1, retrieved using the HV, HXZ, HX,

HY, and HZ methods 83

List of Figures

Figure 2.1: Classification of Profilometers 11

Figure 2.2: The principle of laser triangulation 14

Figure 2.3: The principle of confocal 15

Figure 2.4: Schematic diagram of a point autofocus instrument 16

Figure 2.5: The principle of point autofocus 16

Figure 2.6: Schematic diagram of a focus variation instrument 19

Figure 2.7: Schematic diagram of a phase-shifting interferometer 20

Figure 2.8: Schematic diagram of a digital holographic microscopy 22

Figure 2.9: Schematic diagram of a coherence scanning interferometer 23

Figure 2.10: Keyence confocal displacement sensor 25

Figure 2.11: Chromatic confocal point sensor 26

Figure 3.1: Schematic of Carl Zeiss Axiotron 2 VIS-UV CSM confocal microscope 30

Figure 3.2: Experiment setup for the calibration process 31

Figure 3.3: The relationship between the depth positions and RGB values 34

Figure 3.4: The relationship between the depth positions and HSV 36

Figure 3.5: The calibration curve of the sample (10¢ coin) in HSV space 37

Figure 3.6: The confocal image of the Singapore 10¢ coin 38

Figure 3.7: The surface topography of the Singapore 10¢ coin retrieved using the calibrated confocal microscope 38

Figure 3.8: The surface topography of Singapore 10¢ coin retrieved using Alicona Infinite Focus profilometer 39

Figure 3.9: Confocal microscopic image (20×) of a microfluidic device sample 40

Figure 3.10: The surface topography the microfluidic device retrieved using the calibrated confocal microscope 40

Figure 4.1: The schematic diagram for the prototype system 43

Figure 4.2: Tungsten-Halogen Lamp Spectral Distribution at different color

temperatures 44Figure 4.3: The Kohler’s illumination scheme 45Figure 4.4: Layout of the prototype illumination system (ZEMAX software)

46Figure 4.5: ZEMAX simulation results for the prototype illumination system

(the image of lamp filament is perfectly diffused) 47Figure 4.6: ZEMAX simulation results for a bad illumination system (the

image of lamp filament is clearly seen) 47Figure 4.7: Spectral response curves of a 3CCD camera with 5 channels which

is not suitable for our prototype system 50Figure 4.8: Six spectral response curves that are ideal for the camera used in

confocal 3D profilometry 51Figure 4.9: Spectral response curves of the color camera used in our prototype

52Figure 4.10: The design of the Nipkow Disk 55Figure 4.11: Cross-section side view of the Nipkow disk (not according to

scale) 57Figure 4.12: The reflectance on the top surface of the Nipkow disk 57Figure 4.13: The image (sample: a mirror) when the Nipkow disk is not

spinning 58Figure 4.14: Comparison between the wedge prisms with different wedge

angle 59Figure 4.15: The prototype of the high speed confocal 3D profilometer 60Figure 4.16: A schematics (left) and a photo (right) of the prototype system 61Figure 4.17: The user graphic interfaces of the calibration curve extraction

program 61Figure 4.18: The user graphic interfaces of the surface height retrieval

program 62Figure 5.1: Surface topography of the sample (Rubert 511E) retrieved using a

commercial profilometer, ContourGT-K 68Figure 5.2: Calibrated RGB curves of the prototype system for the Rubert

511E sample 69

Figure 5.3: A confocal image of the Rubert 511E sample captured by the

prototype system 70Figure 5.4: The 3D profile of the sample retrieved using the discrete point

technique with the step size of 1 µm 71Figure 5.5: The 2D profile of the sample retrieved using the discrete point

technique with the step size of 1 µm 71Figure 5.6: The 3D profile of the sample retrieved using the discrete point

technique with the step size of 0.5 µm 72Figure 5.7: The 2D profile of the sample retrieved using the discrete point

technique with the step size of 0.5 µm 72Figure 5.8: The relation between a measuring point and calibration points in a

color information space 73Figure 5.9: The 3D profile of the sample retrieved using the vector projection

technique 75Figure 5.10: The 2D profile of the sample retrieved using the vector projection

technique 76Figure 5.11: The diamond turned step sample has nine step heights 77Figure 5.12: The calibrated RGB curves for the diamond turned step sample

78Figure 5.13: The calibrated HSV curves for the diamond turned step sample

79Figure 5.14: The calibrated XYZ curves for the diamond turned step sample

79Figure 5.15: A confocal image (Image 1) for the diamond turned step sample

81Figure 5.16: A confocal image (Image 7) for the diamond turned step sample

81Figure 5.17: The 2D profile of the Image 1 retrieved using the XYZ method

81Figure 5.18: The 2D profile of the Image 7 retrieved using the XYZ method

82Figure 5.19: The 2D profile of the Image 7 retrieved using the XZ method 83Figure 5.20: The calibration curve of the XZ method shows that the height

z=13 may be retrieved wrongly as z=5 85

Figure 5.21: The calibration curve of the XYZ method 85Figure 5.22: The 2D profile of the Image 7 retrieved using the HX method 86Figure 5.23: The 2D profile of the Image 7 retrieved using the HY method 86Figure 5.24: The calibration curve for the HX method prone to height retrieval

mistake for some regions 87Figure 5.25: The calibration curve for the HY method prone to height retrieval

mistake for some regions 87Figure 5.26: The 2D profiles of the Image 7 retrieved using the RGB method

and the XYZ method 89Figure 5.27: The 2D profiles of the Image 7 retrieved using the HSV method

and the XYZ method 89Figure 5.28: The 2D profiles of the Image 7 retrieved using the HS method

and the XYZ method 90Figure 5.29: The 2D profiles of the Image 7 retrieved using the HXZ method

and the XYZ method 90Figure 5.30: The surface on the right hand side is outside the accurate

measurement zone, thus there are many height retrieval mistakes 91Figure 5.31: The in-focus zone can be divided into one accurate measurement

zone and two rough measurement zones (The diamond turned step sample) 91Figure 5.32: The calibration curve of the XYZ method showing accurate

measurement zones, rough measurement zones and out-of-focus zone 92Figure 5.33: The calibration curve of the XYZ method for the Rubert 511E

sample is very similar to those of the diamond turned step sample 93Figure 5.34: The in-focus zone can be divided into one accurate measurement

zone and two rough measurement zones (The Rubert 511E

sample) 94Figure 5.35: A confocal image of the Rubert 511E sample in the rough

measurement zone 1 95Figure 5.36: A 2D profile retrieved by the XYZ method when the sample is at

rough measurement zone 1 95Figure 5.37: A confocal image of the Rubert 511E sample in the rough

measurement zone 2 95

Figure 5.38: A 2D profile retrieved by the XYZ method when the sample is at

rough measurement zone 2 95Figure 5.39: Two sets of the calibrated XYZ curves are overlapped and used to

retrieved sample surface heights 97Figure 5.40: Six sets of color information are present in the overlapped of two

sets of the calibrated XYZ curves 97Figure 5.41: A profile retrieved from two images where the sample surfaces

are in the rough measurement zones (Experiment Set 1) 98Figure 5.42: A profile retrieved from two images where the sample surfaces

are in the rough measurement zones (Experiment Set 2) 98Figure 5.43: A profile retrieved from two images where the sample surfaces

are in the rough measurement zones (Experiment Set 3) 98Figure 5.44: A profile retrieved from two images where the sample surfaces

are in the rough measurement zones (Experiment Set 4) 98Figure 5.45: The way to increase the measurement range (Strategy I) 99Figure 5.46: The way to increase the measurement range (Strategy II) 100Figure 5.47: Two sets of calibrated XYZ curves are overlapped and used for

height retrieval when the Strategy II is applied 101Figure 5.48: Two sets of calibrated XYZ curves are overlapped and used for

height retrieval when the Strategy II is applied 102Figure 5.49: The first confocal image of the microfluidic device sample 103Figure 5.50: The second confocal image of the microfluidic device sample 104Figure 5.51: The surface topography of the microfluidic device measured by

the prototype system 104

3CCD three charge coupled device

AFM atomic force microscope

B the blue channel (or blue component in RGB color space)

CMM coordinate measurement machine

CSI coherence scanning interferometer

DHM digital holographic microscope

EFC electrostatic force microscope

FWHM full width half maximum

G the green channel (or green component in RGB color space)

H hue, one of the component in HSV color space

HSV the hue, saturation, and value (color space)

MEMS microelectromechanical system

MFM magnetic force microscope

PSI phase shifting interferometer

R the red channel (or red component in RGB color space) RGB the red, green, and blue (color space)

S saturation, one of the component in HSV color space

SNR signal to noise ratio

SEM scanning electronic microscope

SPM scanning probe microscope

SThM scanning thermal microscope

STM scanning tunneling microscope

TEM transmission electronic microscope

V value (lightness), one of the component in HSV color space

X the ratio of red (R) to the sum of R, G, and B

XYZ the ratios of R, G, and B to the sum of R, G, and B

Y the ratio of green (G) to the sum of R, G, and B

Z the ratio of blue (B) to the sum of R, G, and B

a challenge to all existing metrology equipment [9] For example, coordinate measurement machine (CMM) with a contact type of stylus used to be a very popular measuring machine in the manufacturing industry [10] After that, noncontact type three-dimensional (3D) measurement instruments [11] have become popular Nowadays, 3D profilometers [12] have become key metrology instruments for many current state-of-art manufacturing industries

Most manufactured parts rely on some form of control of their surface features The surface is usually the feature on a component or device that interacts with the environment in which the component is housed or the device operates [13] In a miniaturized world, the surface topography has dominant influence on the functional features of a part For example, the surface topography is dominant influence on friction, wear mechanism and adhesive property of miniature contact parts The surface topography of parts also determines how much actual area of contact between two surfaces, which has dominant influence to the electrical and thermal conductivity in a miniature device The surface topography also affects how much light interacts with the part or how the part looks and feels As the parts to be manufactured become smaller and smaller, the surface texture or surface roughness has become an important factor in determining the satisfactory performance of a work piece [14] Thus, industry nowadays has high demand on surface topography measurement tools, i.e 3D profilometers

Due to the great demand of 3D profilometer spans many fields in science and industry, various profilometry techniques have been studied and developed by scientists and engineers [15] However each technique has its advantages and limitations For example, the structure lighting [16] and the Moiré [17] techniques are more suitable for vision inspection of large surface, but its measurement resolution is relatively low (>3µm) Vertical scanning interferometry [18] has high measurement resolution (e.g 3nm to sub-nanometer) and large measurement range (can be >10mm), but it requires step

by step scanning resulting in slow measurement speed The scanner generally

moves at speed of about 5µm per second Phase shifting interferometry (PSI)

[19] has very high resolution in direction, but its measurement range in

z-direction is small (e.g when the wavelength of 632nm is used, it can achieve resolution ≈0.1nm, and measurement range ≈160nm) Atomic force microscope (AFM) [20] has very high resolution (typically 0.1nm [21]), but its measurement speed is slow (>1 minute per image) and field of view is very

small (typically 70 - 150 µm in x and y axis)

Among of the 3D profilometry technologies, confocal profilometry [22] is one of the most promising technologies because of its unique advantages, i.e the vertical resolution and measurement range of a confocal system can be customized to suit wide range of applications For example, the confocal point sensors that designed for high resolution applications can achieve the measurement resolution of 0.01µm and the confocal point sensors that designed for large range measurements can cover the measurement range

of 20mm.In addition, specimens do not require any special surface preparation process such as gold sputtering Unlike phase shifting interferometric techniques [23], confocal techniques do not encounter phase wrapping problem When compare to triangulation [24] and pattern projection [25] confocal techniques do not suffer the drawbacks such as occlusion, which brings shadings due to two optical axes; and also multiple reflection, which caused by specular parts on the surface

However the main drawback for the currently available commercial confocal profilometer is its low measurement speed Commercial confocal

profilometers such as Nanovea ST400 [26], Solarius™ LaserScan [27],

Nanofocus µsurf explorer [28], FRT MicroSpy® Profile [29], STIL Micromesure [30], comprising of a confocal point sensor plus the X and Y

scanning mechanism These profilometers require a few minutes to complete a 3D measurement of an area (e.g 659 × 494 measurement points) The low measurement speed has limited the usefulness of the confocal profilometer

Besides high precision and high accuracy, high measurement speed is another of the requirements for the metrology instrument by today industry Industry requires high speed metrology instrument so that it can be used to perform in-situ and real time measurement With a high speed confocal system

it is probably more feasible to do 100% inspection of items rather than just sampling This will then offer better quality control for parts that are produced

The objective of this research is to design and develop a high speed confocal 3D profilometer, which is able to measure the dimension of various types of structure in microfluidic device The profilometer not only must be able to measure the simple dimension such as width and the depth of microfluidic channels, it also must be able to measure dimension of various types of structure in microfluidic, e.g the miniature pillars structure Miniature pillars [31] are a kind of 3D feature of microfluidic device, widely used as various filters The diameter of the pillars can be as small as 5 µm for example The height of pillars can be as high as 50 µm and the distances between one pillar and other pillars maybe for example 15 µm The surface of microfluidic devices is normally smooth, specular and transparent In mass production, all

of these dimensions are required to be in-situ monitored at the production speed It is a challenge for existing metrology technology not merely because

of the large measurement range and small resolution, but also because it requires high measurement speed

Most existing commercial profilometers are not suitable to measure the mentioned pillar structure of microfluidic device For examples, the profilometers which are based on phase-shifting interferometry principle are not able to measure surface with pillar structure because of phase unwrapping problem due to large step height (i.e pillar) The focus variation [32] profilometers are not able to measure the surface due to the transparent surface

of the microfluidic device Profilometers based on point-autofocus [33] are not suitable because light path may be blocked by the pillars during scanning process Profilometers that used either triangulation, structure light or pattern projection are not able to measure transparent object Both vertical scanning interferometry and confocal technique require mechanical scan, thus its measurement speed is slow

High speed confocal 3D profilometer not only can be used to measure 3D features of microfluidic devices, it also has many applications For example, it can also be used to perform quality inspection in aluminum and steel cold metal rolling, as well as measurement of inkwell volume during large area printing (roll-to-roll) [34] manufacturing process High speed confocal 3D profilometer also can be used to investigate aspheric element of micro lens manufactured via lithographic techniques or injection molding In

the field of precision machining, the 3D profilometer also can be used to determine the quality of cutting tools and the quality of surfaces after milling

In the field of material science and metallography, the 3D profilometer can be used to characterize tribology of engineering surface in manufacture of polymer, foams, solder, textile, etc The 3D profilometer also can be used to measure surface roughness of lead frame and ceramic packages, as well as geometry and other critical parameters in flip chip and advanced packaging process in semiconductor industry In the field of medical implant, 3D profilometer is needed to monitor the roughness of the bone implants, as well

as the shape and surface quality of the contact area

1.2 Contributions

In this work, a high speed confocal 3D profilometer prototype system has been designed, developed and tested The prototype system has be used to measure surface topography of a Rubert Precision Reference Specimen

number 511E, a diamond turned step sample, and a microfluidic devices with

micro pillar structures Prior to the design and the development of high speed confocal 3D profilometer prototype system, experiments have been conducted

to convert a commercial confocal microscope into a confocal profilometer Experiment results show that confocal profilometry is a suitable candidate to measure micro pillar structures of microfluidic device The experiment also proved that high speed confocal 3D profilometry is feasible by combining spinning Nipkow disk and chromatic confocal technique

One of the key issues for the prototype system is low signal to noise ratio (SNR) To address this issue, a pair of wedge prism is added, one above and one beneath the Nipkow disk, and the Nipkow disk surface is tilted to a particular angle Due to the constraint of time, budget and manpower, the design of the prototype system is kept as simple as possible Although the prototype system is not perfect, it has achieved the main of objectives, i.e it has proven that the concept of high speed confocal 3D profilometer is feasible

Currently, all the microscopy system components of our prototype, i.e tube lens, chromatic aberration lens, and microscope objective, are purchased from Carl Zeiss [35] The microscopy system can produce chromatic aberration of about 55 µm for visible light (400nm to 700nm wavelength) In order to increase the measurement range, the microscopy system should have larger chromatic aberration So, four new sets of microscopy systems have been designed and analyzed using ZEMAX [36] software These four sets of microscopy systems can produce chromatic aberration of 116 µm, 171 µm,

146 µm and 148 µm respectively However, these components are not fabricated and not tested due to time and budget constraint

One of the main contributions of this research is the development of the surface height retrieval algorithm for the confocal profilometer system Various surface height retrieval methods such as HSV, HS, XYZ, XZ, HXZ,

HX, HY and HZ method have been tested, analyzed and compared Experimental results show that the XYZ method is the best method because the method is the most accurate, precise and robust method compare to other

methods Initially, the surface height of a pixel is determined in discrete step based on the discrete point technique After that, the Vector Projection technique has been developed to replace the discrete point technique By using the Vector Projection technique, the resolution of the surface height has been improved drastically

It was found that the in-focus zone of a confocal profilometry system can be subdivided into two rough and one accurate measurement zones If the sample surface is inside the accurate measurement zone, one confocal image is enough to retrieve the height of the pixels If the sample surface is outside the accurate measurement zone, then two confocal images, one from the rough measurement zone 1 (blue zone), and another one from the rough measurement zone 2 (red zone), are required in order to retrieve the surface heights with errors of less than 0.5 µm

Another contribution of this research is the development of the multiple-image height retrieval scheme With the multiple-image height retrieval scheme, the measurement range of the system can be increased For the samples consist of single step structure or pillar structures, the sample surfaces to be measured can be divided into two different height categories One surface is called the top surface as its height is at higher position while another surface is refer to as the base surface This kind of sample can be measured by capturing just two confocal images In the first image, the base surface has to be inside the accurate measurement zone Then, the sample is

moved downwards with a known distance so that the top surface is inside the accurate measurement zone in the second image

1.3 Organization

This thesis is organized as follows Chapter 2 provides an overview of different types of optical profilometry techniques, it advantages and limitations Chapter 3 describes the preliminary experimental study which has been done prior to the design of the high speed confocal 3D profilometer Chapter 4 describes the design and development of the high speed confocal 3D profilometer prototype system Experimental results of the prototype system are presented and discussed in Chapter 5 Finally, conclusion and future work

are given in chapter 6

There are various types of profilometers The classification of profilometers is as shown in Figure 2.1 First of all, the profilometers are divided into two main categories, i.e conventional and unconventional [38]; Conventional profilometers do not involve the application of the quantum physics knowledge while unconventional profilometers do Vertical resolution

of unconventional profilometers can reach sub-angstrom and angstrom level while conventional profilometers cannot

Figure 2.1: Classification of Profilometers

Conventional profilometers can be classified into two main groups, i.e contacting profilometers and non-contacting profilometers Contacting profilometers [39] have a stylus touching the surface which is being measured, while non-contact profilometers do not touch the surface during measurement Thus, one of the main advantages of the non-contact profilometers over the contacting profilometers is that it will not damage the surface Non-contact profilometers can be optical or non-optical Optical profilometers based on different optical principles such as triangulation, focus detection, confocal, interferometry or scattering [40] Non-optical profilometry uses the mean other than visible light such as capacitance (capacitive) [41], back pressure (pneumatic) [42], ultrasonic [43] to measure the surface topography

Unconventional profilometry includes particle beam microscopy, scanning probe microscopy [44] and X-rays [45] Examples of particle beam microscopy are scanning electron microscope (SEM), transmission electron microscopes (TEM) [46] Scanning probe microscope (SPM) is a serial measurement device, which uses a nano-scale probe to trace the surface of the sample based on local physical interaction Different types of SPMs are based

on different types of physical interaction between the probe and the surface Scanning tunneling microscope (STM) is based on the quantum-mechanical tunneling effect Atomic force microscope (AFM) uses inter-atomic or intermolecular forces The physical interaction between the probe of a magnetic force microscope (MFM) and specimen surface is magnetic force Electrostatic force microscope (EFC) is based on electrostatic force Scanning thermal microscope (SThM) uses thermocouple probe STM and AFM are two most popular SPMs [47]

2.2 An Overview of Optical Profilometers

In this section, the most common optical profilometry techniques used

by commercial profilometers will be introduced briefly The techniques can be classified into three main categories, i.e point-wise techniques, whole filed techniques, and area-integrating techniques Point-wise techniques measure surface topography point by point via scanning a light spot across the surface, akin to the operation of stylus instrument Point-wise techniques also called scanning optical techniques Three famous point-wise techniques are triangulation, confocal and point auto-focus The major drawback of point-

wise techniques is depended of the scanning rate and number of measurement points

Unlike point-wise techniques, whole-field techniques do not need lateral scanning (also called transverse scanning) because an area of surface topography data is acquired simultaneously Whole-field techniques are also known as full-filed techniques or areal optical techniques Some popular whole-field techniques include focus variation, phase-shifting interferometry, digital holographic microscopy, coherence scanning interferometry, pattern projection, etc

Area-integrating techniques are also referred to as scattering techniques [48] The techniques sample the scattering light from the specimen surface over an area, and then retrieve statistical parameter of the surface based on scattering models built according to various theories The techniques

do not measure the actual peaks and valleys of the surface texture; rather they measure some aspect of the surface height distribution The scattering instruments can be very fast and relatively immune to environmental disturbance

2.3 Point-wise Optical Techniques

2.3.1Triangulation

The principle of triangulation (or laser triangulation) [49] is shown in Figure 2.2 Light from a laser source is projected onto the surface and the light

scatters An imaging lens focuses the scattered light to a spot on a coupled device (CCD) line array or position-sensitive detector As the topography of the surface changes, the position of the focusing beam spot will change accordingly Thus, the height of the surface can be determined

charge-Figure 2.2: The principle of laser triangulation [12]

Profilometers that use triangulation principle suffer from a number of disadvantages First, the laser beam size varies throughout the vertical range because the laser beam is focused through the measuring range The size of the spot will act as an averaging filter near the beginning and end of measuring ranges as the beam size is larger there Second, the measurement fails if the line of sight between laser, surface and detector is blocked (e.g staircase-like specimen)

2.3.2 Confocal

Confocal profilometry is based on the principle of confocal microscopy as shown in Figure 2.3 A confocal microscope uses point illumination and a pinhole in an optically conjugate plane in front of the

detector to eliminate out-of-focus signal Therefore only the light very close to the focal plane can be detected The point light source is imaged onto the specimen Then the light is reflected by the specimen and is imaged onto a conjugate plane in front of the detector At conjugate plane, there is an aperture to filter out out-of-focus signal If the specimen surface is at the focal plane, the image of the point light source on the specimen surface and the image on the detector will be very sharp and has high intensity If the specimen surface is out of focal plane, both images are defocus and only a small part of light is detected by the detector

Figure 2.3: The principle of confocal

2.3.3 Point Autofocus

The schematic diagram of a point autofocus instrument is shown in Figure 2.4 The principle of point autofocus operation is illustrated in Figure 2.5 A laser beam with high focusing properties is generally used as the light source The input beam passes through one side of the objective, and the reflected beam passes through the opposite side of the objective after focusing

on the specimen surface at the centre of optical axis This forms a spot image

on the autofocus sensor after passing through an imaging lens

Figure 2.4: Schematic diagram of a point autofocus instrument [13]

Figure 2.5: The principle of point autofocus [13]

Figure 2.5A shows the in-focus state where specimen is in focus Figure 2.5B shows the defocus state where the surface of the specimen is

below the focus position and the in-focus position The spot image position on

the autofocus sensor displaced accordingly (W) When the autofocus sensor

detects the displacement of the spot image, it feedback the information to autofocus mechanism that adjust the position of objective in order to regain the in-focus status, as shown in Figure 2.5C The height of the specimen

surface can be obtained from the moving distance of the objective, Z2

The main drawback of the profilometers that use the point autofocus principle is that it requires a longer measuring time than other non-contact measurement methods This is because the point autofocus instrument not only requires lateral scanning, but also vertical scanning by moving the objective back to in-focus status The uneven optical intensity within the laser spot (speckle) could generate focal shift errors

The advantages of the point autofocus method are that it almost immune to the surface reflectance properties, it has relatively high resolution, and able to measure the surface with the slope angles greater than the half aperture angle of the objective

2.4 Whole-Field Optical Techniques

2.4.1 Focus Variation

Focus variation requires an optical system with very small depth of focus and a mechanism for vertical (or axial) scanning Figure 2.6 shows the schematic diagram of a focus variation instrument Light emerging from a

white light source is focused onto the specimen via objective Light reflected back from the specimen is gathered by a light-sensitive sensor Due to the small depth of field of the optics, only small regions of the object are sharply imaged The optical system is moved vertically along the optical axis while continuously capture the data from the surface This ensures that each region

of the object is sharply focused Algorithms analyze the variation of focus along the vertical axis in order to convert the acquired data into 3D topographic information

Since focus variation depends on analyzing the variation of focus, it can only measure surfaces where the focus varies sufficiently during the vertical scanning process Thus it is difficult or not able to measure transparent specimens or ultra-smooth surface The vertical resolution of a focus variation instrument depends on the chosen objective and can be as low as 10 nm The vertical scan range depends on the working distance of the objective and ranges from a few millimeters to approximately 20 mm or more

Figure 2.6: Schematic diagram of a focus variation instrument [13]

2.4.2 Phase Shifting Interferometry

Phase-shifting interferometry (PSI) [50-51] uses the interference of two coherent lights to measurement the topographic information of a surface Figure 2.7 shows the schematic diagram of a PSI instrument Light from a coherent light source (i.e laser) is split into two beams by a beam splitter One beam is directed to a reference path where there is a flat and smooth mirror at the end of the path The beam is reflected from the mirror is named as reference beam Another beam is directed to the specimen The light beam reflected back from the specimen surface is called object beam After that the reference beam and object beam are combined and form an interference image

on a CCD camera The interference image is called interferogram or fringe

pattern If the distance of the reference path changes, the phase difference between the reference beam and the object beam will shift too During measurement, several known phase-shifts are introduced and produce changes

in the fringe pattern The fringe patterns are analyzed in computer using phase-shifting algorithm to obtain phase maps Before running phase-unwrapping algorithms, the phase maps can only have the value between 0 to 2π Phase-unwrapping algorithm suppresses the 2π ambiguity and allows the phase maps to have values larger than 2π Finally the vertical height data are deduced from the phase maps

Figure 2.7: Schematic diagram of a phase-shifting interferometer [13]

PSI instrument can have many different configurations Two most common configurations are Mirau configuration and Linnik configuration Most PSI instruments usually require that the height difference between adjacent points on a surface should not more than λ/4 One of the major

drawbacks of phase-shifting interferometry techniques is that the measurement results are always wrapped in a 2π area This means that these methods have difficulty in measuring surfaces with discontinuities such as large steps and isolated area Phase-unwrapping usually are complicated process and take out

a lot of time To avoid phase-unwrapping process, the range of PSI instrument

is usually limited to one fringe, or approximately half the central wavelength

of the light source

2.4.3 Digital Holographic Microscopy

A digital holographic microscope (DHM) [52-54] is an interferometric microscope very similar to PSI, but with a small angle between the propagation directions of the object beam and reference beam as shown is Figure 2.8 The image acquired by DHM is called digital hologram Unlike PSI, the interference of the object beam and the reference beam of the DHM is off-axis Therefore, the acquired digital hologram consists of a spatial amplitude modulation with successive constructive and destructive interference fringes Thus only one digital hologram is needed to reconstruct the phase map

In most of DHM instrument, the image of the object formed by the microscopic objective is not focused on the camera DHM is equipped with a numerical wavefront propagation algorithm that uses numerical optics to increase the depth of field or compensate for optical aberration DHM has a similar resolution to PSI and is limited in range to half the central wavelength

of the light source when a single wavelength is used However,

dual-wavelength DHM [55-56] or multiple dual-wavelength DHM [57] allows the vertical range to be increased to several micrometers

Figure 2.8: Schematic diagram of a digital holographic microscopy [13]

2.4.4 Coherence Scanning Interferometry

The configuration of a coherence scanning interferometer (CSI) [58] is similar to that of a phase-shifting interferometer except that in CSI a broadband (white light) or extended (many independent point sources) source

is used These are low coherence light sources and the light emerge has short coherence length CSI is also known as vertical scanning interferometer [59], white light interferometer [60], vertical scanning white light interferometry, or

white light scanning interferometry Figure 2.9 shows a schematic diagram of

a coherence scanning interferometer

Figure 2.9: Schematic diagram of a coherence scanning interferometer [13]

As in PSI, the light in CSI is split into object beam and reference beam and then recombined for interference Due to low coherence of the source, the optical path length to the sample and reference must be almost identical, for interference to be observed The camera continues to capture images as the optical path is varied in the vertical direction (z axis) Each pixel of the camera measures the intensity of the light and fringe envelope obtained can be used to calculate the height position of the surface